The Vegas Golden Knights erased a three-goal deficit to defeat the Colorado Avalanche 5-3 at T-Mobile Arena, taking a commanding 3-0 lead in the Western Conference Final.
Colorado led 3-0 after the first period. Vegas scored five unanswered goals across the final two periods to complete the comeback. The Avalanche now face elimination, their 121-point regular season on the brink of ending without a Stanley Cup Final appearance.
This wasn't a goaltending collapse. Carter Hart stopped 31 of 34 shots for Vegas. Scott Wedgewood made 18 saves on 22 shots for Colorado. The difference came down to territorial control after the opening frame and whether Colorado's depth forwards could sustain offensive pressure once Vegas adjusted defensively.
Parker Kelly has scored 21 goals on 13.0 expected goals this season. Jack Drury has 10 on 13.2 expected. Both were on the ice when Vegas completed their comeback. Kelly is running 8.0 goals above expected, per MoneyPuck. When you combine elite finishing with poor underlying play-driving, you get exactly what happened in this game: early production masking structural problems that Vegas exploited once they tightened defensively.
Colorado's 121 points in the regular season were built on elite talent at the top of the roster and unsustainable depth scoring that was always going to regress. It just happened to regress in Game 3 of a conference final.
The Pattern That Killed Colorado
Gabriel Landeskog scored at 3:21 of the first period. Nazem Kadri made it 2-0 at 7:03. Drury added a shorthanded goal at 13:15, assisted by Kelly and Devon Toews. That shorthanded tally will appear in the box score as a positive contribution from the depth forwards. The underlying numbers tell a different story.
Drury carries a 48% on-ice expected goals share at five-on-five this season. Kelly sits at 43%. These aren't players driving offense through sustained pressure. They're players whose shooting percentages have been covering for the fact that they lose the territorial battle nearly every shift. When both are on the ice together, the Avalanche spend more time defending than attacking, hoping to convert on the occasional transition chance they generate.
That works in the regular season across 82 games when variance evens out. It doesn't work in a playoff game where the opposition adjusts after getting punched in the mouth, tightens gap control, and forces you to create quality chances through sustained offensive zone time rather than rush opportunities.
Vegas scored 19 seconds into the second period on the power play. Mark Stone finished a feed from Mitchell Marner and Tomas Hertl, cutting the deficit to 3-1. William Karlsson made it 3-2 at 4:05. Keegan Kolesar tied it at 12:46. Hertl gave Vegas the lead at 8:21 of the third period, and Brett Howden finished the scoring at 19:01.
The box score shows Colorado outshot Vegas 34-22. The underlying story is that once Vegas adjusted their neutral zone structure and began activating their defensemen more aggressively without completely opening up, Colorado's depth lines couldn't generate enough offensive zone time to protect the lead.
When Shooting Percentage Meets Playoff Reality
Nathan MacKinnon posted a 65% on-ice expected goals share this season with 12.4 goals above expected. Martin Necas posted 64% with 11.4 goals above expected. Landeskog posted 63%. These players both finish at elite levels and create chances at elite rates. When they're on the ice, Colorado dominates territorially.
When Kelly and Drury are on the ice, Colorado's on-ice expected goals share drops into the low 40s, and the team spends most shifts defending. The gap between those two realities didn't matter much over 82 games. In a conference final game where Vegas can deploy Jack Eichel, Stone, and Hertl against those depth lines and suffocate them in the neutral zone, that gap becomes fatal.
Kelly's 8.0 goals above expected is exactly the kind of number that screams regression. Drury sitting 3.2 goals below expected despite playing sheltered minutes suggests he's not even generating quality looks when he does shoot. The combination of poor chance creation and shooting percentages that were always going to normalize created a ticking time bomb that went off in the second and third periods of this game.
Ross Colton played 8:29, registered four hits, and generated one shot attempt. That's a physical presence. It's not offense. It's not the kind of contribution that helps you protect a three-goal lead against a team that can layer their forecheck and force you into low-percentage plays.
The Structural Reality
Colorado's bottom-six forwards posted a combined 43% on-ice expected goals share this season and were on the ice for four of Vegas' five goals in this game. That's not a coincidence. That's a roster construction issue masked by elite finishing that was never sustainable.
Wedgewood stopped 18 of 22 shots. The reason Colorado lost isn't that he should have stopped one or two more. The reason Colorado lost is that Vegas controlled play for long stretches after the first period, and Colorado's depth forwards couldn't generate enough sustained offensive zone pressure to relieve that territorial disadvantage.
When you're getting outchanced at a 43% rate during your shifts, you're not giving your goaltender much help. When your bottom-six can't even break even territorially against Vegas' depth lines, your stars have to be perfect for 60 minutes. MacKinnon, Necas, and Landeskog were excellent. They just couldn't carry the entire load once Vegas tightened up structurally.
The Avalanche finished the regular season with a 55-16-11 record and 121 points. Their goal differential suggested dominance. The underlying process suggested a team whose depth scoring was living on shooting percentages that wouldn't survive playoff defensive intensity. This game provided the evidence.
What Vegas Understood
Vegas didn't win this through brilliance. They won it by recognizing that if they could survive the first period, Colorado's depth lines would eventually struggle to sustain offensive pressure. Stone, Marner, and Hertl combined for six points. Howden and Kolesar added goals. Vegas forced Colorado's bottom-six to try to generate offense in a structured defensive environment, and those lines couldn't do it.
The question now is whether Colorado can adjust with their backs against the wall, trailing 3-0 in the series. Can the coaching staff shelter the depth lines more effectively? Can they find a way to extract more offensive zone time from players who have been losing the territorial battle all season?
If Kelly and Drury continue posting sub-45% on-ice expected goals shares, Colorado won't win another game this spring. The stars will create chances, the depth will give them back, and the Avalanche will exit wondering why their 121-point regular season didn't translate. The answer is in the advanced stats: depth scoring built on shooting percentage rather than chance creation doesn't survive when the games matter most.